This project is an experimental neuroevolution simulator in Scratch. At the moment, it works, but does not do anything practical. Soon, it may learn to play platformer games. For now, there is not much to see, unless you want to modify it yourself or read the advanced explanation.
An advanced explanation: The current objective is to move towards the right as quickly as possible. It creates a list of 10 random movements, 10 times, finds which one is the best, and evolves it in the next generation. Every 10 trials are separated into a generation, and the best one moves on to the next generation. Each genome is a separate modification of the best genome of the previous generation. Fitness is how far the dot has moved on the screen. The "genbest" list is the best genome from the previous generation. The "bestscript" list is the best genome from the current generation. The "queue" list is the current genome being tested. This algorithm can be modified to play platform games.
